Pandas data frame: groupby and plot with two different columns

Question

I am a super beginner for Python. Long story short, I want to groupby with one column, apply one function to one column, apply another function to another column, and plot the results(the first column to the x-axis, the second column to the y-axis).

I have a pandas data frame df which contains many columns. Two columns of them are tour_id and tour_distance .

tour_id    tour_distance    
      A               10
      A               10
      A               10
      A               10
      B               20
      B               20
      C               40
      C               40
      C               40
      C               40
      C               40
      :                :
      :                :

Since I assume that the longer tour_distance becomes, the more rows each tour_id has, I want to plot a histogram of tour_distance vs row counts in each group of tour_id .

Question 1: what's the simplest solution for this groupby and plot problem?

Question 2: how can I improve my failed attempt?

My attempt: I thought it would be easier to make a new data frame like this.

tour_id    tour_distance  row_counts
      A               10           3
      B               20           2
      C               40           5
      :                :           :

In this way I can use matplotlib and do like this,

import matplotlib.pyplot as plt
x = df.tour_distance
y = df.row_counts
plt.bar(x,y)

However, I can't make this data frame.

df_tour_distance = df.groupby('tour_id').tour_distance.head(1)
df_tour_distance = pd.DataFrame(df_tour_distance)
df_size = df.groupby('tour_id').tour_distance.size()
df_size = pd.DataFrame(df_size)
df = pd.merge(df_size, df_tour_distance, on='tour_id')

>>> KeyError: 'tour_id'

This also failed:

g = df.groupby('tour_id')
result = g.agg({'Count':lambda x:x.size(), 
            'tour_distance_grouped':lambda x:x.head(1)})
result

>>> KeyError: 'Count'

Answer 1

The problem in your code is that once you groupby tour_id , it becomes index. You have to specify as_index=False or use reset_index() in order to use it. Also, you do not need to find a series and then merge it back.

You need:

g = df.groupby(['tour_id', 'tour_distance']).size().reset_index(name='count')
plt.bar(g['tour_id'],g['count'])

Output:

Answer 2

Could be implemented somewhat easier:

import pandas as pd

tour_id = ['A']*4+['B']*2+['C']*5
tour_distance = [10]*4+[20]*2+[40]*5

df = pd.DataFrame({'tour_id': tour_id, 'tour_distance': tour_distance})
df = df.set_index('tour_id')

df2 = pd.DataFrame()
df2['tour_distance'] = df.groupby('tour_id')['tour_distance'].head(1)
df2['row_counts'] = df.groupby('tour_id').count()
print(df2)

Result:

         tour_distance  row_counts
tour_id                           
A                   10           4
B                   20           2
C                   40           5